Cooperative Game Theory:

Characteristic Functions, Allocations, Marginal Contribution

Adam Brandenburger

Version 01/04/07

1 Introduction
Game theory is divided into two branches, called the non-cooperative and co-
operative branches. The payo¤ matrices we have been looking at so far belong
to the non-cooperative branch. Now, we are going to look at the cooperative
branch.

The two branches of game theory di¤er in how they formalize interdepen-
dence among the players. In the non-cooperative theory, a game is a detailed
model of all the moves available to the players. By contrast, the cooperative
theory abstracts away from this level of detail, and describes only the outcomes
that result when the players come together in di¤erent combinations.

Though standard, the terms non-cooperative and cooperative game theory

are perhaps unfortunate. They might suggest that there is no place for cooper-
ation in the former and no place for con‡ict, competition etc. in the latter. In
fact, neither is the case. One part of the non-cooperative theory (the theory of
repeated games) studies the possibility of cooperation in ongoing relationships.
And the cooperative theory embodies not just cooperation among players, but
also competition in a particularly strong, unfettered form.

The non-cooperative theory might be better termed procedural game the-

ory, the cooperative theory combinatorial game theory. This would indicate
the real distinction between the two branches of the subject, namely that the
…rst speci…es various actions that are available to the players while the second
describes the outcomes that result when the players come together in di¤erent
combinations. This is an important analytical distinction which should become
clear as we proceed.

The idea behind cooperative game theory has been expressed this way:

With the assistance of Amanda Friedenberg. co o p -01 -04 -07

 “Cooperative theory starts with a formalization of games that
abstracts away altogether from procedures and . . . concentrates,
instead, on the possibilities for agreement. . . . There are sev-
eral reasons that explain why cooperative games came to be treated
separately. One is that when one does build negotiation and en-
forcement procedures explicitly into the model, then the results of a
non-cooperative analysis depend very strongly on the precise form of
the procedures, on the order of making o¤ers and counter-o¤ers and
so on. This may be appropriate in voting situations in which precise
rules of parliamentary order prevail, where a good strategist can in-
deed carry the day. But problems of negotiation are usually more
amorphous; it is di¢ cult to pin down just what the procedures are.
More fundamentally, there is a feeling that procedures are not re-
ally all that relevant; that it is the possibilities for coalition forming,
promising and threatening that are decisive, rather than whose turn
it is to speak. . . . Detail distracts attention from essentials. Some
things are seen better from a distance; the Roman camps around
Metzada are indiscernible when one is in them, but easily visible
from the top of the mountain.”
(Aumann, R., “Game Theory,” in Eatwell, J., Milgate, M., and
P. Newman, The New Palgrave, New York, Norton, 1989, pp.8-9.)

2 De…nition of a Cooperative Game

A cooperative game consists of two elements: (i) a set of players, and (ii) a
characteristic function specifying the value created by di¤erent subsets of the
players in the game. Formally, let N = f1; 2; : : : ; ng be the (…nite) set of players,
and let i, where i runs from 1 through n, index the di¤erent members of N .
The characteristic function is a function, denoted v, that associates with every
subset S of N , a number, denoted v(S). The number v(S) is interpreted as the
value created when the members of S come together and interact. In sum, a
cooperative game is a pair (N; v), where N is a …nite set and v is a function
mapping subsets of N to numbers.

Example 1 As a simple example of a cooperative game, consider the following

set-up. There are three players, so N = f1; 2; 3g. Think of player 1 as a seller,
and players 2 and 3 as two potential buyers. Player 1 has a single unit to sell,
at a cost of $4. Each buyer is interested in buying at most one unit. Player
2 has a willingness-to-pay of $9 for player 1’s product, while player 3 has a
willingness-to-pay of $11 for player 1’s product. The game is depicted in Figure
1 below.

2
 WtP $11

WtP $9

Cost $4 Cost $4

Figure 1

We de…ne the characteristic function v for this game as follows:

v (f1; 2g) = $9 $4 = $5;

v (f1; 3g) = $11 $4 = $7;
v (f2; 3g) = $0;
v (f1g) = v (f2g) = v (f3g) = $0;
v (f1; 2; 3g) = $7:

This de…nition of the function v is pretty intuitive. If players 1 and 2 come

together and transact, their total gain is the di¤erence between the buyer’s
willingness-to-pay and the seller’s cost, namely $5. Likewise, if players 1 and 3
come together, their total gain is the again the di¤erence between willingness-
to-pay and cost, which is now $7. Players 2 and 3 cannot create any value by
coming together; each is looking for the seller, not another buyer. Next, no
player can create value on his or her own, since no transaction can then take
place. Finally, note that v (f1; 2; 3g) is set equal to $7, not $5 + $7 = $12. The
reason is that the player 1 has only one unit to sell and so, even though there are
two buyers in the set f1; 2; 3g, player 1 can transact with only one of them. It is
a modeling choice–but the natural one–to suppose that in this situation player
1 transacts with the buyer with the higher willingness-to-pay, namely player 3.
That is the reason for setting v (f1; 2; 3g) equal to $7 rather than $5.1

3 Marginal Contribution
Given a cooperative game (N; v), the quantity v(N ) speci…es the overall amount
of value created. (In the example above, this quantity was v (f1; 2; 3g) = $7.)
1 For more discussion of this last, somewhat subtle point, see “Value-Based Business Strat-

egy,” by Adam Brandenburger and Harborne Stuart, Journal of Economics & Management
Strategy, Spring 1996, Section 7.1 (“Unrestricted Bargaining”), pp.18-19.

3
An important question is then: How is this overall value divided up among the
various players? (Referring again to the example above, the question becomes:
How does the $7 of value get split among the three players?)

The intuitive answer is that bargaining among the players in the game de-
termines the division of overall value v(N ). This bargaining in typically ‘many-
on-many.’ Sellers can try to play one buyer o¤ against another, buyers can try
to do the same with sellers. Intuitively, a player’s ‘power’ in this bargaining
will depend on the extent to which that player needs other players as compared
with the extent to which they need him or her. Brie‡y put, the issue is: Who
needs whom more?

The analytical challenge is to formalize this intuitive line of reasoning. This

is what the concept of marginal contribution does.

To de…ne marginal contribution, a piece of notation is needed: Given the

set of players N and a particular player i, let N nfig denote the subset of N
consisting of all the players except player i.

De…nition 1 The marginal contribution of player i is v(N ) v (N nfig), to

be denoted by MCi .

In words, the marginal contribution of a particular player is the amount by

which the overall value created would shrink if the player in question were to
leave the game.

Example 1 Contd. To practice this de…nition, let us calculate the marginal

contributions of the players in the example above:

MC1 = v (f1; 2; 3g) v (f2; 3g) = $7 $0 = $7;

MC2 = v (f1; 2; 3g) v (f1; 3g) = $7 $7 = $0;
MC3 = v (f1; 2; 3g) v (f1; 2g) = $7 $5 = $2:

We will return to this example once more below, to use these marginal-
contribution numbers to deduce something about the division of the overall
value created in this particular game. Before that, we need to state and justify
a principle about the division of value in a cooperative game. (Later on, we will
examine a stricter principle than this one.)

De…nition 2 Fix a cooperative game (N; v). An allocation is a collection

(x1 ; x2 ; : : : ; xn ) of numbers.

4
 The interpretation is easy: An allocation is simply a division of the overall
value created, and the quantity xi denotes the value received by player i.

De…nition 3 An allocation (x1 ; x2 ; : : : ; xn ) is individually rational if xi

v(fig) for all i.

Pn
De…nition 4 An allocation (x1 ; x2 ; : : : ; xn ) is e¢ cient if i=1 xi = v(N ).

These two de…nitions are quite intuitive. Individual rationality says that a
division of the overall value (i.e. an allocation) must give each player as much
value as that player receives without interacting with the other players. E¢ -
ciency says that all the value that can be created, i.e. the quantity v(N ), is in
fact created.

Unless otherwise noted, all allocations from now on will be assumed to be

individually rational and e¢ cient.

De…nition 5 An (individually rational and e¢ cient) allocation (x1 ; x2 ; : : : ; xn )

satis…es the Marginal-Contribution Principle if xi MCi for all i.

The argument behind the Marginal-Contribution Principle is simple, almost

tautological sounding. First observe that the total value captured by all the
players is v(N ). It therefore follows from the de…nition of MCi that if some
player i were to capture more than MCi , the total value captured by all the
players except i would be less than v (N nfig). But v (N nfig) is the amount of
value that these latter players can create among themselves, without player i.
So, they could do better by coming together without player i, creating v (N nfig)
of value, and dividing this up among themselves. The putative division of value
in which player i captured more than MCi would not hold up.

The Marginal-Contribution Principle, while indeed almost obvious at some

level, turns out to o¤er a single, far-reaching method of analyzing the division
of value in bargaining situations. The power of the principle should become
apparent as we explore some applications.

Example 1 Contd. As a …rst application, return one more time to Exam-

ple 1 above. The overall value created was $7. Let us now use the marginal-
contribution calculations above to deduce something about the division of this
value among players 1, 2, and 3. Since player 2 has zero marginal contribution,
the Marginal-Contribution Principle implies that player 2 won’t capture any
value. Player 3 has a marginal contribution of $2, and so can capture no more

5
than this. This implies that player 1 will capture a minimum of $7 $2 = $5.
The remaining $2 of value will be split somehow between players 1 and 3. The
Marginal-Contribution Principle does not specify how this ‘residual’ bargaining
between the two players will go.

This answer is very intuitive. In this game, the buyers (players 2 and 3) are
in competition for the single unit that the seller (player 1) has on o¤er. The
competition is unequal, however, in that player 3 has a higher willingness-to-pay
than player 2, and is therefore sure to win. The operative question is: On what
terms will player 3 win? The answer is that player 2 will be willing to pay up
to $9 for player 1’s product, so player 3 will have to pay at least that to secure
the product. The e¤ect of competition, then, is to ensure that player 1 receives
a price of at least $9 (hence captures at least $9 $4 = $5 of value); player 3
pays a price of at least $9 (hence captures at most $11 $9 = $2 of value); and
player 2 captures no value. Will the price at which players 1 and 3 transact be
exactly $9? Not necessarily. At this point, the game is e¤ectively a bilateral
negotiation between players 1 and 3, in which player 1 won’t accept less than
$9 and player 3 won’t pay more than $11. The Marginal-Contribution Principle
doesn’t specify how this residual $2 of value will be divided. It allows that
player 3 might pay as little as $9, or as much as $11, or any amount in-between.
A reasonable answer is that the residual $2 will be split evenly, with player 1
capturing a total of $5 + $1 = $6 of value, and player 3 capturing $1 of value
(but other answers within the permissible range are equally acceptable).

We see that the cooperative game-theoretic analysis captures in an exact

fashion the e¤ect of competition among the players in a bargaining situation.
It makes precise the idea that the division of value should somehow re‡ect who
needs whom more. But the analysis is (sensibly) agnostic on where the bar-
gaining over residual value–what remains after competition has been accounted
for–will end up. After all, where this leads would seem to depend on ‘intan-
gibles’such as how skilled di¤erent players are at persuasion, blu¢ ng, holding
out, and so on. These are factors external to the game as described in coopera-
tive theory. Thus, an indeterminacy in the theory at this point is a virtue, not
a vice.